Collect structured data

Use Track actions to collect structured rows into a project Container. Define the Container and fields first, then select their actual definitions in Toolbox.

Build a collection

1. Add a Track Container action and select the destination Container.
2. For a list of page rows, add a locator selecting those rows.
3. Set an explicit maximum iteration count when collecting more than one row.
4. Inside its body, add Track Data Field actions for declared fields.
5. Set each child locator relative to the selected row when appropriate, then Apply each edited action.
6. Play the complete Container subtree and inspect preview output.

Use the Container/field selector proposals and relative preview to check the row family before execution. New field locators use inherited row context; review existing absolute choices rather than silently converting them.

The default Container maximum is one row. While and Repeat On have a separate Toolbox Debug behavior: outer loops advance one iteration per Play while Debug is on. Turn Debug off for a complete multi-page rehearsal, or deliberately advance/restart iterations.

The Container limit is independent of that switch. A matching list of many elements does not automatically mean all rows will be collected. Without a Container locator, the body runs once.

Relative extraction

Use inherited locator context for a child field selected inside the current row. An empty inherited CSS selector targets the row itself. An absolute locator searches the document instead, which can accidentally repeat the same first value for every record.

Use a specific selector such as .price or a.title inside the row. Test against rows with missing and empty values as well as ordinary rows.

Values and formatting

Fields can use a value binding or extraction. Multiple extracted text matches join with a space by default; a supported separator can change this. A template format can use currentValue for the current extraction.

No matching element omits the field. Explicit false, zero, empty text and null remain distinct from absence. A field action needs an enclosing active Container; playing that field alone does not reconstruct the collection context.

A row is completed only after its body finishes normally. Failure, cancellation or Exit discards that unfinished row. Earlier confirmed rows remain.

Preview versus publication

Toolbox always previews collection. It does not publish records, even when the Job’s data-publication setting is enabled. Preview output is not a stored record identifier. The live view is bounded and may retain only recent completed rows from a large test; do not treat its current row count as a full export.

To verify persistence, save the Job, review Push data in effective settings, run a small ordinary Session and inspect Data afterwards, or the selected external destination when one is configured. A configured external destination does not retain a duplicate internal business-data copy. With publication disabled, ordinary collection also does not save business records.

Stop on existing matters only when real publication can check the key. It accepts the current record and stops that Container traversal after encountering an existing key. It does not stop the whole Job and has no existing-record effect in Toolbox preview.